Using Adler-32 as checksum algorithm for linear store (I think it's a bad idea)

[Andrew Stitcher <as...@redhat.com>]

It seems that Adler-32 has bad properties for short sequences of bytes
(say up to 128 bytes). See: https://tools.ietf.org/html/rfc3309 for a
cogent explanation why.

I think that many of the messages stored will be fairly short, or at
least there should be no loss of protection for short messages. So I
think we should hastily change the checksum algotihm before it gets out
into the wild (perhaps to crc-32c like sctp, but this maybe deserves
some study).

Andrew



---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@qpid.apache.org
For additional commands, e-mail: dev-help@qpid.apache.org